Effects of Psychosocial Characteristics of Middle School Students on High School Grades and On-Time Graduation

ACT WORKING PAPER 2015-07

Joann L. Moore, PhD Jason D. Way, PhD Alex Casillas, PhD Jeremy Burrus, PhD Jeff Allen, PhD Mary Ann Hanson, PhD July, 2015

ACT Working Paper Series ACT working papers document preliminary research. The papers are intended to promote discussion and feedback before formal publication. The research does not necessarily reflect the views of ACT.

This form of the manuscript is under review at European Journal of Personality Assessment. It has not yet been accepted for publication.

Joann Moore is a research scientist in Statistical and Applied Research specializing in prediction of secondary and postsecondary outcomes from academic and non-cognitive factors.

Jason Way is a research psychologist in Career Transitions Research. He researches behavioral and psychosocial skills, including their relevance to important academic and work outcomes.

Alex Casillas is a senior research psychologist in Research for Assessment and Learning specializing in assessment design and behavioral predictors of performance and persistence.

Jeremy Burrus is a principal research scientist in the Career Transitions Research department working on the assessment of noncognitive skills.

Jeff Allen is a statistician in the Research division at ACT. He specializes in longitudinal research linking test scores to educational outcomes and student growth models.

Mary Ann Hanson is director of Career Transitions Research. She is involved in research and development for a wide variety of career- and behavior-related solutions across education and work.

Abstract

Research has shown that psychosocial factors (PSFs) have a positive impact on high school

outcomes, including grades and persistence (e.g., Farrington et al., 2012). However, few

longitudinal studies have examined the nature of these relationships. We report on a longitudinal

study of middle school students followed through high school completion. We found that high

school GPA mediates the effects of prior academic achievement, demographics, and most PSFs

on on-time high school graduation. A measure of self-regulation had a significant direct effect

on on-time graduation. The results underscore the importance of PSFs in predicting academic

outcomes, even after accounting for prior achievement and demographics. Implications of

examining PSFs early in students’ academic progression are discussed.

Effects of psychosocial characteristics of middle school students on high school grades and on-time graduation

Although the situation has improved in recent years, on-time high school graduation

remains of major concern in the U.S., with over 20% of entering public high school students

failing to earn a high school diploma in four years (Aud et al., 2013). In some states and

communities, these rates exceed 50% of all entering 9th grade students. In the discussion that

follows, we use the term “dropout” to refer to the situation where a student’s enrollment in high

school is suspended before they earn a diploma. While on-time graduation is not the direct

opposite of dropout, the two outcomes are closely related and the dropout literature is relevant

for understanding on-time graduation.

Over the past decade, research has shown that measuring critical psychosocial factors

(PSFs; e.g., motivation, social engagement, self-regulation) can increase schools’ abilities to

identify and intervene with students at risk of dropout or delayed graduation (e.g., Farrington et

al., 2012; Zins, Bloodworth, Weissberg, & Walberg, 2004). As the dropout process begins well

before students actually decide to leave high school (e.g., Rumberger & Lim, 2008), research

into early indicators of dropout or delayed graduation is needed. Further, research should focus

not only on academic indicators, but psychosocial indicators as well.

Early Warning for Dropout

Research has demonstrated that high school dropout can be predicted well before students

enter high school. For example, in public schools in the Northeastern U.S., an early warning

system to identify 6th grade students at risk to drop out in high school was created (Balfanz,

Herzog, & Mac Iver, 2007; Neild, Balfanz, & Herzog, 2007). Indicators included grades,

attendance, and disciplinary records. Together these indicators identified 60% of the students

who did not graduate within one year of expected graduation. Further, several large-scale

longitudinal studies have found that at ages 2, 6, 12, and 14, future high school dropouts were

much more likely than graduates to have: parents who were divorced or separated; lower

achievement test scores; and, higher scores on measures of behavioral problems (see Heckman,

Humphries, & Kautz, 2014 for a review). This research clearly shows that academic

performance and behavioral indicators measured during middle school can be effective

predictors of on-time high school graduation. Many of these indicators (e.g., attendance,

discipline, behavior problems) are related to PSFs.

Psychosocial Factors (PSFs)

Research has examined the effects of PSFs on academic performance and persistence,

highlighting a range of constructs, including self-efficacy, motivation, locus of control, attitude

toward learning, and persistence (e.g., Grigorenko et al., 2009; Poropat, 2009; Robbins et al.,

2004). Other studies have found that PSFs provide incremental validity over traditional

predictors of academic performance such as achievement tests and demographics (e.g., Casillas

et al, 2012; Zins et al. 2004). Analogous results have been obtained in the postsecondary realm

(e.g., Robbins, Allen, Casillas, Peterson, & Le, 2006), as well as for work outcomes such as job

performance and retention (e.g., Barrick & Mount, 1991; Judge & Ilies, 2002).

One assessment of PSFs is ACT Engage, a low-stakes, self-report instrument designed to

measure ten PSFs related to academic performance and persistence (Casillas et al., 2011), which

can be grouped into three broad behavioral domains: motivation, social engagement, and self-

regulation (see Table 1). Motivation refers to the mechanism by which individuals act on

prescribed behaviors and implement learning activities and/or pursue goals, social engagement

refers to an individual’s skills in engaging the social environment in ways that support and

reinforce his or her learning activities, and self-regulation refers to the ability to manage or

regulate attitudes, behaviors, and feelings that affect students’ receptiveness to, and

implementation of, learning activities (Robbins et al., 2009).

Casillas et al. (2012) conducted a longitudinal study to validate the ACT Engage

instrument for middle school students, using a large (N~5,000), longitudinal dataset of students

followed from middle school through the first year of high school. They found that PSFs

measured by Engage provided incremental validity above achievement tests and prior grades

when predicting early high school grade point average (HSGPA) and that the variance accounted

for was comparable to that of prior grades.

Current Study

This study builds on the Casillas et al. (2012) study, following the same sample after

additional years of data collection to examine effects of PSFs on HSGPA and on-time

graduation. A path model to examine direct and indirect effects of PSFs on intermediate (grades)

and more distal (graduation) outcomes is proposed, with grades serving as a potential mediator

of the relationship between PSFs and on-time graduation (see Figure 1). This allows for the

examination of differential effects of the PSFs and other predictors on both outcomes.

Specifically, the research questions addressed are:

RQ1: Is there evidence of effects of PSFs on HSGPA, after controlling for prior academic

achievement and student and school demographics?

RQ2: Is there evidence of effects of PSFs on on-time high school graduation, after

controlling for prior academic achievement, student and school demographics, and HSGPA?

This question examines whether PSFs have direct effects on on-time graduation and whether

the relationship is mediated by HSGPA.

RQ3: How do effects on on-time high school graduation vary by PSF? This question

examines which PSFs have the largest effects on on-time graduation.

Method

A prospective sample of 4,660 middle-school students from 24 schools in 13 districts

throughout the U.S. completed PSF and standardized achievement tests during the fall of 2006

when most students were in 8th grade. The 24 schools varied with respect to percentage of

students eligible for free or reduced lunch (mean=53%, minimum=29%, maximum=97%) and

percentage of students in underrepresented minority groups (mean=27%, minimum=1%,

maximum=97%). Follow-up data were collected each fall (2007-2011), with districts providing

data annually on GPA, absences, and enrollment status. The last wave of data was collected

after most students would have graduated from high school on time.

Measures

Academic achievement in middle school was measured using ACT Explore, a

standardized achievement test typically taken in 8th or 9th grade, which includes measures of

English, mathematics, reading, and science (ACT, 2013). Subject area scores range from 1-25

and the Composite score is calculated as the mean of the four subject area scores. Explore is

intended for all students in grades 8 and 9 and focuses on the knowledge and skills that are

usually attained by grade 8.

PSFs were measured using ACT Engage Grades 6–9. The instrument contains 97 items

scored using a 6-point Likert-type scale ranging from strongly disagree to strongly agree and

nine yes/no items; the 106 items form 10 scales corresponding to each PSF (see Table 1 for scale

names, definitions, and example items). The ten scales demonstrate moderate to high internal

consistency reliabilities (range α=.81–.90; Mdn α=.87; Casillas et al., 2011).

School demographic information included percent of students eligible for free or reduced

lunches (FRL), percent minority (African American, Hispanic, and American Indian), and size

(defined as the average cohort size across grade levels served by the school). School

demographics were obtained from the Common Core of Data (nces.ed.gov/ccd). Student

demographic variables included gender and indicators for race/ethnicity categorized as white,

African American, Hispanic, and other (American Indian, Asian, two or more races, and other

were combined due to small sample sizes).

Outcomes. Mid-point HSGPA was defined as students’ end-of-10th grade cumulative

GPA. If a student’s 10th grade cumulative GPA was missing, then his/her end-of-9th grade

cumulative GPA was used instead. Students were classified as having graduated high school on

time if they had graduated within four years of starting high school and were classified as

unsuccessful if they had dropped out, were expelled, had earned a General Educational

Development (GED) high school equivalency certificate, or were still enrolled after four years of

starting high school. Only students for whom on-time graduation could be determined were

included in the analyses. Two schools were excluded from the analyses (n=192) because the

final wave of data collection was unavailable. A small number of students (approximately 1%)

were in 7th grade in fall of 2006, and therefore may have been in 12th grade in the fall of 2011

during the last wave of data collection; if such a student was enrolled in fall 2011, they were

excluded from the analysis. Students who were classified as deceased (n=4), home schooled

(n=84), not enrolled–unknown (n=382), or transferred (n=709), were excluded from the analyses

because on-time graduation was undefined or unknown. Overall, 87% of students in the analysis

sample were classified as graduated on time.

Data Cleaning

Records were also excluded if a student did not complete any of the Engage scales (n=9)

or did not take the ACT Explore test (n=557), yielding a total sample size of 2,764 from 21

schools. The analysis sample appears similar to the original sample; average demographic

characteristics (school percent FRL, school and student percent minority, percent male) were

within 5 percentage points of the original sample. In the analysis sample, mean Explore

Composite score was .11 SD units higher, and means for each of the Engage scales were higher

by .02 to .08 SD units compared to the original sample.

Approximately 30% of the sample completed an early form of the Engage assessment

that did not include one of the scales, Relationships with School Personnel. All other predictors

had a missing rate of 10% or less, and missing values were imputed for all predictors. Because

the multiple imputation procedure borrows information from other variables used in the

imputation process, a larger set of predictors than those of interest were included in the

imputation procedure.

Analyses

To address the first research question (RQ1), a multilevel linear model was used to relate

the predictors to HSGPA. To account for variation in grades across schools not explained by the

student and school predictors, a random intercept model was used. The effects of each student-

level predictor were assumed to be the same across schools, thus random slopes were not used.

To address RQ2, multilevel logistic regression was used to relate the predictors to on-time

graduation. Again, to account for school effects, random intercepts were used, but not random

slopes.

The results from the models for HSGPA and on-time graduation can be used to determine

the extent that HSGPA mediates effects on on-time graduation, using Baron and Kenny’s (1986)

approach to testing for mediation. Indirect effects were estimated by the product of the

coefficients (e.g., α×β in Figure 1B; Wright, 1934) and significance was tested using Sobel’s

(1982) approach. Total effects of each predictor were estimated by fitting the model for on-time

graduation without the mediator (HSGPA) (e.g., τ in Figure 1A). RQ3 is addressed by

examining differences in total effects across the ten PSF measures. To facilitate comparisons of

effect size, all predictors except for the student demographics (gender and race/ethnicity

indicators) were standardized prior to modeling.

Results

Table 2 contains descriptive statistics for all variables in the model prior to imputation

and standardization. Mean Explore Composite score was 15.2 (standard deviation 3.2), which is

similar to the U.S. national fall grade 8 mean (15.5) and standard deviation (3.3) (ACT, 2013).

Fifty-one percent of students were female, 67% white, 9% African American, 12% Hispanic, and

89% spoke English as their primary language.

Table 2 also contains correlations between all study variables after imputing missing

data. All predictors were significantly correlated with both HSGPA and on-time high school

graduation, and HSGPA and on-time high school graduation were correlated with one another

(r=.53). Explore Composite and the Engage scales were more highly correlated with HSGPA

than with on-time graduation. Explore Composite was positively correlated with each of the

Engage scales, and the Engage scales were all positively intercorrelated. The largest

intercorrelation among predictors was observed for two of the PSF scales within the self-

regulation domain (Orderly Conduct and Managing Feelings, r=.64), suggesting that there does

not appear to be serious redundancy in the predictors.

Table 3 presents the results of the regression models and the indirect, direct, and total

effects of all predictors on on-time graduation. Because all but on-time high school graduation

and the student demographic variables were standardized, the beta coefficients can be interpreted

as the increase in HSGPA in standard deviation units–or increase in the log-odds of on-time high

school graduation–corresponding to an increase of one standard deviation in the predictor,

controlling for the other predictors in the model. For the student demographic indicators, the

betas correspond to the increase in the outcome corresponding to demographic group

membership.

To address RQ1, the first set of coefficients corresponds to the model predicting HSGPA.

Higher school percent minority and male status predicted lower HSGPA. Higher Explore

Composite score, Academic Discipline, and Orderly Conduct predict higher HSGPA. A small

but significant negative relationship was found between Commitment and HSGPA. Because the

bivariate relationship between Commitment and HSGPA was positive (r=.28), this is likely an

artifact of multicollinearity, also known as suppression.

To address RQ2, the direct effects on on-time graduation were examined. The only

significant predictors in the full model (p < 0.05) were Orderly Conduct and HSGPA. HSGPA

fully mediated the relationships between school percent minority, male status, Explore

Composite, Academic Discipline, and Commitment on on-time graduation. Orderly Conduct

remained a significant predictor of on-time graduation, but the coefficient was reduced in

magnitude, indicating partial mediation. Optimism and School Climate were marginally

significant (p < 0.1), with the coefficient for School Climate suggesting a negative effect.

Indirect effects were obtained by multiplying the coefficients from the model predicting

HSGPA with the HSGPA-on-time graduation coefficient in the full (direct effects) model.

Significant indirect effects were found for school percent minority and male status, Explore

Composite score, Academic Discipline, Commitment, and Orderly Conduct. This was expected

because the same variables were significant predictors of HSGPA.

The final set of coefficients in Table 3 corresponds to the total effects, or the effects of

the predictors on the outcome without the presence of the mediator (GPA) in the model.

Because multilevel linear and logistic models were used, the indirect and direct effects do not

sum to the total effects. The indirect effects of school percent minority, male status, and

Commitment were significant, but the total effects were not. The total effects of Optimism and

Thinking Before Acting were significant, but the indirect effects were not. Both the indirect and

total effects of Family Attitude and School Climate were marginally significant (p < 0.1).

Interestingly, the total effect of Thinking Before Acting was negative, despite a positive bivariate

relationship with the outcome.

To address RQ3, we compared the total effects across the different PSF measures. The

largest effects were observed for Orderly Conduct (0.372), Academic Discipline (0.340),

Optimism (0.175), and Thinking Before Acting (-0.176). For the other six PSFs, there was no

evidence of additional effects on on-time graduation. Thus, among the constructs measured by

the Engage assessment, there is evidence of effects from the motivation and self-regulation

domains, but no additional effects from the social engagement domain. The negative estimate

for Thinking Before Acting was not expected and, as mentioned earlier, may be the product of

multicollinearity. In a model for on-time graduation that excluded all other PSFs but included

mid-point GPA, the coefficient for Thinking Before Acting was positive and not significant. The

predictors with significant effects on on-time graduation are summarized in Figure 2

(Commitment and Thinking Before Acting are not included because their effects were smaller and

negative, contrary to theory).

Discussion

This study showed that PSFs measured in middle school predict both high school grades

and on-time high school graduation. While observational longitudinal studies do not bear

evidence of causality, the results suggest that PSFs measured in middle school are related to

academic performance and persistence in high school. They also highlight the value of using

PSFs to supplement traditional predictors of academic outcomes. Identifying students’ strengths

and needs in terms of their PSFs allows teachers and administrators to approach student success

from a more nuanced perspective, rather than solely focusing on academic performance. Given

that student risk factors can be identified at an early age, it may be possible to avert future

negative outcomes. Further, schools can connect students to targeted resources and interventions

that can help them improve in areas needing development and thus be less likely to drop out or

delay graduation.

RQ1 examined which PSFs predicted mid-point HSGPA, and thus had indirect effects on

on-time graduation. Academic Discipline and Orderly Conduct were important predictors of

HSGPA. These results show that, for middle school students, being motivated to complete

school work and behaving appropriately are precursors to later success in high school. Male

status was negatively related to HSGPA, and as expected, grade 8 academic achievement

(Explore Composite) was positively related.

RQ2 examined which PSFs have direct effects on on-time graduation. Mid-point GPA

fully mediated the relationships found between on-time high school graduation and

demographics, prior academic achievement, and most of the prior PSFs, and partially mediated

the relationship between high school graduation and Orderly Conduct. The impact of

demographics, prior academic performance, and most PSFs on high school graduation can be

explained by their effects on HSGPA, suggesting that lowering risk for poor academic

performance may also lower risk for dropout or delayed graduation. Orderly Conduct was the

only PSF to retain a direct effect on graduation after accounting for the mediating effect of

HSGPA. This is in line with previous research showing that poor grades and poor conduct are

two pathways that lead students to drop out of school (e.g., Rumberger & Lim, 2008). The direct

effect of Optimism was positive and nearly statistically significant, suggesting that a hopeful

outlook on the future helps students persist in high school.

RQ3 examined differences in total effects on on-time graduation, across the ten PSF

scales. Five of the ten PSFs had significant direct, indirect, or total effects on at least one of the

outcomes. Ordering by the magnitude of their total effect on graduation, the most important

PSFs were Orderly Conduct, Academic Discipline, Optimism, and Family Attitude.

Limitations

One limitation of this study is that the participating schools were located mainly in the

Midwestern and Southern United States, and were not nationally representative. However, the

students included represented a wide range of individual differences. Additionally, some error

may have been introduced to the model due to the imputation of missing values. On-time

graduation status was unknown for students who transferred out of the study districts or who

were no longer enrolled for reasons unknown, and were thus excluded from analysis. Also, the

study relied on self-reports of the PSFs. The effects of PSFs on high school outcomes may have

been stronger if the self-report measures were supplemented by observer ratings (e.g., teacher

behavioral ratings), or if effect estimates were corrected for measurement error.

While the study included a diverse set of predictors of HSGPA and graduation, the results

do not prove that the predictors caused the outcomes. This limitation is important because it

calls into question whether interventions can affect the outcome (on-time graduation) by

improving PSFs (e.g., Orderly Conduct).

Future Research

Additional research would benefit from incorporating other measures not included in this

study, such as absenteeism and other academic behaviors (e.g., homework completion, failed

classes, disciplinary actions). Future studies also could use alternative measurement methods

(e.g., other reports, behavior diaries) to help validate the types of behaviors that are predictive of

high school outcomes.

Future research also could incorporate measures of PSFs at later points in time into a

similar research design, and examine the extent to which students’ PSFs change between middle

school and high school and whether the mediating relationship between these factors and

HSGPA holds steady throughout high school. In addition, it would be useful to systematically

explore how interventions implemented based on prior assessment of PSFs can impact student

outcomes.

Conclusion

In a paper cited in the introduction of this study, Neild, Balfanz, and Herzog (2007)

report that they could predict a 6th grade student’s future chance of dropping out of high school

with 75% accuracy using just a few indicators. Consistent with the current study, one of these

indicators had to do with behavioral conduct. In their paper, Neil and colleagues state that,

“These students are metaphorically waving their hands and asking for help” (p. 28). Measures of

PSFs that predict on-time graduation, such as ACT Engage, can, as such, serve as a mode

through which students can “ask for help” and can be used by educators as part of early warning

systems designed to identify students who may be at risk. Although we are unsure whether

intervening directly on the PSFs measured by ACT Engage will lead to lower dropout (see the

limitation about causality discussed above), at minimum such a tool can signal which students

might require some kind of additional supports (psychosocial or otherwise) to complete high

school successfully. That alone, we believe, makes early assessment of PSFs valuable in applied

settings.

References

ACT. (2013). ACT Explore technical manual. Iowa City: Author. Retrieved from

http://www.act.org/explore/pdf/TechManual.pdf

Aud, S., Wilkinson-Flicker, S., Kristapovich, P., Rathbun, A., Wang, X., & Zhang, J. (2013). The

Condition of Education 2013 (NCES 2013-037). U.S. Department of Education, National

Center for Education Statistics. Washington, DC. Retrieved from

http://nces.ed.gov/pubs2013/2013037.pdf.

Balfanz, R., Herzog, L., & Mac Iver, D.J. (2007). Preventing student disengagement and keeping

students on the graduation path in urban middle-grades schools: Early identification and

effective interventions. Educational Psychologist, 42, 223–235.

Baron, R.M., & Kenny, D.A. (1986). The moderator-mediator variable distinction in social

psychological research: Conceptual, strategic, and statistical considerations. Journal of

Personality and Social Psychology, 51(6), 1173-1182. Retrieved from

http://www.public.asu.edu/~davidpm/classes/psy536/Baron.pdf

Barrick, M.R. & Mount M.K. (1991). The big five personality dimensions and job performance:

A meta-analysis. Personnel Psychology, 44, 1-26.

Casillas, A., Allen, J., Kuo, Y.-L., Pappas, S., Hanson, M.A., & Robbins, S.B. (2011).

Development and validation of ENGAGE Grades 6–9. Iowa City: ACT Research

Reports, ACT, Inc.

Casillas, A., Robbins, S., Allen, J., Kuo, Y., Hanson, M.A., & Schmeiser, C. (2012). Predicting

early academic failure in high school from prior academic achievement, psychosocial

characteristics, and behavior. Journal of Educational Psychology, 104, 407-420.

Farrington, C.A., Roderick, M., Allensworth, E., Nagaoka, J., Keyes, T.S., Johnson, D.W., &

Beechum, N.O. (2012). Teaching adolescents to become learners. The role of

noncognitive factors in shaping school performance: A critical literature review.

Chicago: University of Chicago Consortium on Chicago School Research.

Grigorenko, E.L., Jarvin, L., Diffley III, R., Goodyear, J., Shanahan, E.J., & Sternberg, R.J.

(2009). Are SSATs and GPA enough? A theory-based approach to predicting academic

success in secondary school. Journal of Educational Psychology, 101, 964-981.

Heckman, J.J., Humphries, J.E., & Kautz, T. (2014). The Myth of Achievement Tests. Chicago,

IL: The University of Chicago Press.

Judge, T.A. & Ilies, R. (2002). Relationship of personality to performance motivation: A meta-

analytic review. Journal of Applied Psychology, 87, 797-807.

Neild, R.C., Balfanz, R., & Herzog, L. (2007). An early warning system. Educational

Leadership, 65, 28-33.

Poropat, A.E. (2009). A meta-analysis of the five-factor model of personality and academic

performance. Psychological Bulletin, 135, 322-338.

Robbins, S.B., Allen, J., Casillas, A., Peterson, C., & Le, H. (2006). Unraveling the differential

effects of motivational and skills, social, and self-management measures from traditional

predictors of college outcomes. Journal of Educational Psychology, 98, 598–616.

Robbins, S.B., Lauver, K., Le, H., Davis, D., Langley, R., & Carlstrom, A. (2004). Do

psychosocial and study skill factors predict college outcomes? A meta-analysis.

Psychological Bulletin, 130, 261–288.

Robbins, S.B., Oh, I., Le, H., & Button, C. (2009). Intervention effects on college performance

and retention as mediated by motivational, emotional, and social control factors:

Integrated meta-analytic path analyses. Journal of Applied Psychology, 94, 1163-1184.

Rumberger, R.W., & Lim, S.A. (2008). Why students drop out of school: A review of 25 years of

research. Santa Barbara, CA: California Dropout Research Project. Retrieved from

http://cdrp.ucsb.edu/dropouts/ pubs_reports.htm#15

Sobel, M.E. (1982). Asymptotic confidence intervals for indirect effects in structural equation

models. Sociological Methodology, 13, 290-312.

Wright, S. (1934). The method of path coefficients. The Annals of Mathematical Statistics, 5(3),

161-215.

Zins, J.E., Bloodworth, M.R., Weissberg, R.P., & Walberg, H.J. (2004). The scientific base

linking emotional learning to student success and academic outcomes. In J.E. Zins, R.P.

Weissberg, M.C. Wang, & H.J. Walberg (Eds.), Building academic success on social and

emotional learning: What does the research say? (pp. 3-22). New York, NY: Teachers

College Press.

Figure 1. Relationships tested among PSFs, academic achievement, demographics, and high school outcomes.

Figure 2. Effects of predictors on on-time graduation.

Table 1. Engage Grades 6-9 Domains, Scale Names, Definitions, and Sample Items

Domain Scale Name Definition Sample Item

Motivation Academic Discipline

Degree to which a student is hardworking and conscientious as evidenced by the amount of effort invested into completing schoolwork.

I turn in my assignments on time.

Commitment to School

Commitment to stay in school and obtain a high school diploma.

I am committed to graduating from high school.

Optimism A hopeful outlook about the future in spite of difficulties or challenges.

I am confident that everything will turn out all right.

Social Engagement

Family Attitude toward Education

Positive family attitude regarding the value of education.

My family supports my efforts in school.

Family Involvement

Family involvement in a student’s school life and activities.

I talk to my family about schoolwork.

Relationships with School Personnel

The extent to which students relate to school personnel as part of their connection to school.

Adults at my school understand my point of view.

School Safety Climate

School qualities related to students’ perception of security at school.

I feel safe at school.

Self-Regulation

Managing Feelings

Tendency to manage duration and intensity of negative feelings (e.g., anger, sadness, embarrassment) and to find appropriate ways to express feelings.

I would walk away if someone wanted to fight me.

Orderly Conduct Tendency to behave appropriately in class and avoid disciplinary action.

I have been sent to the principal’s office for misbehaving.

Thinking before Acting

Tendency to think about the consequences of one’s actions before acting.

I think about what might happen before I act.

Table 2. Correlations and Descriptive Statistics.

Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

1) School % FRL —

2) School % Minority .44 —

3) School Size -.65 .11 —

4) Male -.01 -.02 .01 —

5) White -.22 -.43 -.04 -.02 —

6) Black .27 .31 -.16 .02 -.45 —

7) Hispanic .02 .21 .15 -.01 -.54 -.12 —

8) Other Race/Ethnicity .05 .11 .03 .03 -.46 -.10 -.12 —

9) Explore Composite -.23 -.11 .18 -.06 .24 -.22 -.09 -.05 —

10) Academic Discipline -.13 -.11 .07 -.17 .13 -.03 -.10 -.06 .26 .90

11) Commitment to Sch. -.12 -.07 .07 -.18 .08 .04 -.09 -.06 .23 .52 .84

12) Family Attitude -.09 -.02 .07 -.07 .06 .01 -.06 -.03 .21 .45 .61 .85

13) Family Involvement -.12 -.10 .04 -.06 .11 .02 -.10 -.07 .14 .52 .47 .57 .86

14) Managing Feelings -.13 -.07 .11 -.23 .15 -.11 -.05 -.07 .20 .52 .28 .23 .39 .90

15) Optimism -.08 -.06 .04 -.09 .05 .06 -.06 -.07 .15 .52 .50 .43 .55 .39 .89

16) Orderly Conduct -.20 -.15 .13 -.23 .20 -.12 -.10 -.07 .28 .56 .30 .24 .34 .64 .31 .82

17) Relat. w/ Sch. Pers. -.19 -.15 .07 -.10 .18 -.06 -.11 -.09 .14 .51 .41 .36 .59 .46 .56 .39 .89

18) School Climate -.18 -.22 .02 -.06 .14 -.06 -.07 -.07 .08 .35 .31 .28 .37 .37 .43 .36 .59 .82

19) Think. Bef. Acting -.06 .03 .06 -.10 .03 .02 -.01 -.04 .15 .50 .27 .25 .39 .57 .42 .49 .41 .31 .86

20) 10th Grade GPA -.14 -.21 .04 -.18 .26 -.15 -.18 -.05 .58 .45 .28 .27 .27 .32 .26 .42 .28 .18 .24 — 21) On-time HS Grad. -.15 -.10 .12 -.10 .13 -.10 -.08 -.03 .29 .27 .23 .21 .19 .18 .19 .26 .17 .10 .12 .53 —

N 2764 2764 2764 2758 2761 2761 2761 2761 2751 2762 2763 2761 2762 2761 2762 2463 1940 2759 2759 2725 2764

Min 0.3 0 39 0 0 0 0 0 5.0 10.0 16.0 10.0 10.0 10.0 10.0 10.0 10.0 10.0 10.0 0 0

Max 1.0 1.0 425 1.0 1.0 1.0 1.0 1.0 25.0 60.0 60.0 60.0 60.0 60.0 60.0 60.0 60.0 60.0 60.0 4.3 1.0

Mean 0.5 0.3 246 0.5 0.7 0.1 0.1 0.0 15.2 48.7 56.7 55.8 47.3 39.1 48.5 44.9 40.2 42.2 39.4 2.7 .87

SD 0.1 0.2 96 0.5 0.5 0.3 0.3 0.3 3.2 8.8 5.5 6.3 10.1 11.8 9.1 14.7 10.0 9.3 10.4 0.9 .30 Note. Correlations >.04 are significant at p < .05. Correlations > .20 are in bold. Cronbach’s alphas for Engage scales are on the diagonal.

Table 3. Indirect, Direct, and Total Effects for the Model.

10th Grade GPA On-time High School Graduation Indirect Effects Direct Effects Total Effects Effect Beta SE P-value Beta SE P-value Beta SE P-value Beta SE P-value Intercept 0.027 0.108 0.805 3.904* 0.574 <.0001 2.784* 0.491 <.0001 School demographics School % FRL 0.075 0.064 0.244 0.158 0.136 0.245 -0.222 0.272 0.413 0.154 0.279 0.582 School % Minority -0.101* 0.045 0.024 -0.214* 0.095 0.025 0.089 0.188 0.634 -0.189 0.191 0.324 School Size -0.011 0.055 0.836 -0.024 0.117 0.836 0.120 0.241 0.619 0.285 0.246 0.247 Student demographics Male -0.170* 0.029 <.0001 -0.360* 0.064 <.0001 0.052 0.161 0.747 -0.251 0.137 0.066 White 0.110 0.104 0.290 0.232 0.220 0.291 -0.385 0.549 0.484 0.048 0.470 0.919 Black 0.000 0.112 0.998 -0.001 0.238 0.998 -0.407 0.576 0.480 -0.123 0.493 0.803 Hispanic -0.114 0.108 0.294 -0.241 0.230 0.295 -0.290 0.568 0.610 -0.340 0.486 0.485 Other Race/Ethnicity 0.079 0.110 0.473 0.167 0.233 0.473 -0.106 0.584 0.856 0.046 0.498 0.926 Prior academic achievement Explore Composite 0.468* 0.015 <.0001 0.990* 0.067 <.0001 0.130 0.107 0.228 0.764* 0.085 <.0001 Psychosocial factors Academic Discipline 0.232* 0.021 <.0001 0.491* 0.052 <.0001 0.034 0.101 0.737 0.340* 0.087 <.0001 Commitment -0.042* 0.019 0.028 -0.089* 0.041 0.029 0.074 0.084 0.381 -0.020 0.070 0.773 Family Attitude 0.036 0.019 0.062 0.075 0.041 0.064 0.078 0.085 0.361 0.136 0.073 0.061 Family Involvement 0.019 0.020 0.348 0.040 0.043 0.348 0.040 0.103 0.697 0.082 0.087 0.347 Managing Feelings -0.003 0.020 0.892 -0.006 0.043 0.892 0.064 0.104 0.537 0.017 0.092 0.852 Optimism 0.016 0.019 0.407 0.033 0.040 0.408 0.184 0.098 0.060 0.175* 0.082 0.033 Orderly Conduct 0.121* 0.020 <.0001 0.257* 0.045 <.0001 0.217* 0.101 0.032 0.372* 0.087 <.0001 Relat. w/ Sch. Pers. 0.033 0.021 0.107 0.071 0.044 0.109 -0.014 0.114 0.902 0.040 0.098 0.685 School Climate -0.033 0.018 0.067 -0.070 0.038 0.069 -0.166 0.096 0.084 -0.160 0.082 0.052 Think. Bef. Acting -0.024 0.018 0.186 -0.051 0.039 0.188 -0.101 0.095 0.287 -0.176* 0.083 0.035 10th grade GPA 2.117* 0.126 <.0001 Note. * Significant at p < 0.05